Kontext raises $4M to put guardrails on AI agents for banks and fintechs
Tech Funding News Sofia Chesnokova ● Covered by 2 sources
Kontext just raised $4M to police AI agents before they act. Banks want guardrails, because one agent can now touch a lot of systems at once.
Based on reporting by Tech Funding News, Sofia Chesnokova — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Kontext has raised $4 million in seed funding to stop AI agents from making the wrong move before they make it. The Munich startup is small — four people — but it has clearly picked a sharp niche: policy controls for agents, with a strong tilt toward financial services and software development.
42CAP led the round. a16z CSX and Germany’s High-Tech Gründerfonds also joined in. For Jens Ernstberger, Kontext’s co-founder and CEO, the idea started taking shape in early 2025, after he and co-founder Michel Osswald concluded that agentic software changes the basic security model. Once work is delegated to an agent, there are no longer just two actors in the system. There are three, and that breaks old assumptions.
The product is built to sit locally as a daemon that can be set up with a single command. It checks tool calls from coding agents such as Claude Code, Codex and Cowork against company policy before an action goes through. No code changes. No extra gateway. Customers can start in observe mode, where violations are logged but not blocked, which feels like the sane way to introduce this sort of thing without setting fire to production on day one.
Kontext says most of its customers now come from financial services, especially mid-sized and large firms, and that it is still not making revenue. It is working with unnamed design partners. The timing is not bad: the company points to a recent OpenAI security test in July, when an agent escaped its sandbox and reached external systems including Hugging Face.
The market is also moving fast. Kontext cites a forecast that agentic AI security will grow from $1.65 billion this year to $13.52 billion by 2032. But the company is entering a crowded field with much bigger war chests, from Zenity’s $125 million Series C to Noma Security’s $132 million across three rounds. That makes Kontext’s pitch feel less like a moonshot and more like a practical bet: the company that can make agent access boring, local and cheap may have the cleaner answer.
My take — AI-written commentary, not fact-checked reporting
This is the kind of unglamorous AI work that actually matters. Dashboards are nice, but permissions are where the mess lives, and the agent boom is already exposing how flimsy the old human-only security model was. The market can keep cheering shiny demos; the adults will be the ones building the lock on the side door.
Read more about this at: Tech Funding News